18.

06 Quiz 3 Solutions May 8, 2010 Profes­

sor Strang

Your PRINTED name is: 1.

Your recitation number is 2.
3.

1. (40 points) Suppose u is a unit vector in Rn , so uT u = 1. This problem is about the n by n

symmetric matrix H = I − 2u uT .

(a) Show directly that H 2 = I. Since H = H T , we now know that H is not only

symmetric but also .

Solution Explicitly, we find H 2 = (I − 2uuT )2 = I 2 − 4uuT + 4uuuT uuT (2 points):

since uT u = 1, H 2 = I (3 points). Since H = H T , we also have H T H = 1, implying

that H is an orthogonal (or unitary) matrix.

(b) One eigenvector of H is u itself. Find the corresponding eigenvalue.

Solution
Since Hu = (I − 2uuT )u = u − 2uuT u = u − 2u = −u, λ = −1.

(c) If v is any vector perpendicular to u, show that v is an eigenvector of H and find the

eigenvalue. With all these eigenvectors v, that eigenvalue must be repeated how many

times? Is H diagonalizable? Why or why not?

Solution For any vector v orthogonal to u (i.e. uT v = 0), we have Hv = (I−2uuT )v =

v − 2uuT v = v, so the associated λ is 1. The orthogonal complement to the space

spanned by u has dimension n−1, so there is a basis of (n−1) orthonormal eigenvectors

with this eigenvalue. Adding in the eigenvector u, we find that H is diagonalizable.

(d) Find the diagonal entries H11 and Hii in terms of u1 , . . . , un . Add up H11 + . . . + Hnn and

separately add up the eigenvalues of H.

Solution Since ith diagonal entry of uuT is u2i , the i diagonal entry of H is Hii = 1−2u2i

(3 points). Summing these together gives ni=1 Hii = n − 2 ni=1 u2i = n − 2 (3 points).
� �

�
Adding up the eigenvalues of H also gives λi = (1) − 1 + (n − 1)(1) = n − 2 (4 points).
2. (30 points) Suppose A is a positive definite symmetric n by n matrix.

(a) How do you know that A−1 is also positive definite? (We know A−1 is symmetric. I just

had an e-mail from the International Monetary Fund with this question.)

Solution Since a matrix is positive-definite if and only if all its eigenvalues are positive

(5 points), and since the eigenvalues of A−1 are simply the inverses of the eigenvalues

of A, A−1 is also positive definite (the inverse of a positive number is positive) (5

points).

(b) Suppose Q is any orthogonal n by n matrix. How do you know that Q A QT = Q A Q−1

is positive definite? Write down which test you are using.

Solution Using the energy text (xT Ax > 0 for nonzero x), we find that xT QAQT x =

(QT x)T A(QT x) > 0 for all nonzero x as well (since Q is invertible). Using the positive

eigenvalue test, since A is similar to QAQ−1 and similar matrices have the same

eigenvalues, QAQ−1 also has all positive eigenvalues. (5 points for test, 5 points for

application)

(c) Show that the block matrix

⎡ ⎤
A A

B =
⎣

⎢ ⎥
⎦

A A
is positive semidefinite. How do you know B is not positive definite?

Solution First, since B is singular, it cannot be positive definite (it has eigenvalues

of 0). However, the pivots of B are the pivots of A in the first n rows followed by 0s

in the remaining rows, so by the pivot test, B is still semi-definite. Similarly, the first

n upper-left determinants of B are the same as those of A, while the remaining ones

are 0s, giving another proof. Finally, given a nonzero vector

⎡
⎤

⎢ x
⎥
u =
⎣
⎦

where x and y are vectors in Rn , one has uT Bu = (x+y)T A(x+y) which is nonnegative

(and zero when x + y = 0).

3. (30 points) This question is about the matrix

⎡ ⎤
0 −1

A =
⎣

⎢ ⎥
⎦

4 0
.

(a) Find its eigenvalues and eigenvectors.

⎡
⎤

⎢ 2
⎥
Write the vector u(0) =
⎣
⎦
as a combination of those eigenvectors.

0
Solution Since det(A − λI) = λ2 + 4, the eigenvalues are 2i, −2i (4 points). Two

associated eigenvectors are [1 − 2i]T , [1 2i]T , though there are many other choices

(4 points). u(0) is just the sum of these two vectors (2 points).

du
(b) Solve the equation = Au starting with the same vector u(0) at time t = 0.
dt

In other words: the solution u(t) is what combination of the eigenvectors of A?

Solution One simply adds in factors of eλt to each term, giving

⎡ ⎤ ⎡ ⎤
⎢ 1
⎥ −2it ⎢
1
⎥
u(t) = e
2it ⎣
⎦
+ e ⎣
⎦

−2i 2i

(c) Find the 3 matrices in the Singular Value Decomposition A = U Σ V T in two steps.

–First, compute V and Σ using the matrix AT A.

–Second, find the (orthonormal) columns of U.

Solution Note that AT A = V ΣT U T U ΣV T = V Σ2 V T , so the diagonal entries of Σ

are simply the positive roots of the eigenvalues of

⎡ ⎤⎡ ⎤ ⎡ ⎤
⎢ 0 4
⎥ ⎢ 0 −1
⎥ ⎢ 16 0
⎥
AT A = ⎣
⎦⎣ ⎦
=
⎣
⎦

−1 0 4 0 0 1

i.e. σ1 = 4, σ2 = 1. Since AT A is already diagonal, V is the identity matrix. The

columns of U should satisfy Au1 = σ1 v1 , Au2 = σ2 v2 : by inspection, one obtains

⎡ ⎤ ⎡ ⎤ ⎡ ⎤
⎢ 0
⎥ ⎢ −1
⎥ ⎢ 0 −1
⎥
u1 = ⎣
⎦
, u2 = ⎣
⎦
, U
=
⎣
⎦

1 0 1 0

MIT OpenCourseWare
http://ocw.mit.edu

18.06 Linear Algebra

Spring 2010

For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.